OpenAI's GPT-6 Sol costs a fifth of Astra per token, and Luna a hundredth
Sam Altman says the pair beats the 5.6 models at half the price per token, and OpenAI's own docs still send new users to Astra first.
OpenAI put two more models in the GPT-6 family on September 22, 2026. GPT-6 Sol costs $2.00 per million input tokens and $10.00 per million output tokens, and GPT-6 Luna costs $0.10 and $0.50.
Those prices are on OpenAI's own model pages. GPT-6 Astra, the flagship, sits at $10 and $50 on the same index, so Sol runs at a fifth of it and Luna at a hundredth.
- GPT-6 Sol
- $2.00 input, $10.00 output, per 1M tokens
- GPT-6 Luna
- $0.10 input, $0.50 output, per 1M tokens
- GPT-6 Astra
- $10 / Input MTok, $50 / Output MTok
- Context window
- 1,050,000
- Max output
- 128,000 tokens
- Knowledge cutoff
- Apr 20, 2026 for Sol, May 18, 2026 for Luna
What the pages say each one is for
Sol is the coding and agent model, "Built to power complex coding and agentic workflows". Luna is "our most efficient model for focused, high-volume tasks". Both read text and images, both write text, and both carry the same 1,050,000-token context window as Astra.
The knowledge cutoffs are the odd part. Luna, the cheapest of the three, is current to May 18, 2026, while Astra stops at April 30 and Sol at April 20. Nothing on either page explains why the newest reading went to the smallest model.
Sol and Luna both take reasoning.effort set to none, which turns the thinking off and, on Chat Completions, is the only setting that allows function calling. Astra's row in the same index runs from low to max with no none on it.
What Altman said about the price
Altman posted about the pair from his own account at 2:26 pm Eastern, comparing them to the 5.6 models they follow.
GPT-6 Sol and Luna are big improvements on intelligence, alignment, work output, coding, computer use, and more over their 5.6-family predecessors. They are also half the price per token, and even less per task!
Two minutes later he moved the comparison off tokens and onto whole tasks, which is the number a company running an agent actually pays.
Especially compared by per-task pricing, which is the metric that should matter, I don't think there is anything competitive anywhere in the market. We want people to be able to use tons of AI; it is important to being able to explore this new renaissance in front of us.
The rest of the bill
Token prices aren't the whole of it. Cached input is billed at 10% of the uncached rate and cache writes at 1.25 times it, batch and flex work at half price, and fast mode doubles whatever you'd otherwise pay. Long prompts cost more per token than short ones.
Prompts with more than 272K input tokens are priced at 2x input and cache rates and 1.5x output for the full request.
For all the price cutting, OpenAI's model index still sends a newcomer to the expensive one: "If you're not sure where to start, use GPT-6 Astra". Sol is offered to balance intelligence and cost, Luna for high-volume work.
Neither of the new models is on the free API tier, which both rate-limit tables mark "Not supported". A Tier 1 key gets 500 requests a minute on either, with a batch queue of 1,500,000 tokens on Sol and 5,000,000 on Luna.
More on OpenAI
- OpenAI model release dates: o3, GPT-5.5 and GPT-6 Astra, Sol and Luna in orderSeptember 23, 2026
- “You exceeded your current quota, please check your plan and billing details.” (OpenAI API)September 23, 2026
- “You’ve hit your usage limit.” in Codex: what each version of the message meansSeptember 23, 2026
- “stream disconnected before completion” in Codex: what drops the stream and what to trySeptember 23, 2026